An assumption becomes useful when it can be challenged
A business assumption is something your plan relies on. A testable hypothesis states a particular claim precisely enough that an observation could count against it. It should identify the relevant people, situation, behavior, and conditions without quietly combining several different questions.
“Cafés need our service” is too broad to guide a test. It might mean owners dislike their current photographs, will book a package, or will reorder. Those are different claims. Strategyzer’s hypothesis guidance distinguishes testable, precise, and discrete statements. Our examples below apply that principle to a small service experiment.
A business experiment is not automatically a controlled scientific study. A clear hypothesis makes interpretation more disciplined, but a small selected sample still has limits.
Split the belief before choosing a test
| Separate claim | What to observe | What it does not establish |
|---|---|---|
| Image updates cause a recurring workflow problem | Recent accounts of the process and consequences | That owners will buy this package |
| Eligible owners will accept a $160 package | Acceptance of the actual scope, price, and dates | Reliable delivery or repeat demand |
| One package can be delivered in 2.5 hours | Measured delivery, travel, and revision time | That more customers can be reached affordably |
Do not run interviews and then declare all three claims validated because people described a problem. Match each claim to evidence that can actually address it.
Write the statement and the decision rule separately
Use this original writing prompt: “For [specific group] in [situation], when offered [clear conditions], we expect to observe [one behavior] within [review period].” This is a prompt, not a guarantee that every valid hypothesis must follow the same sentence structure.
Claim: owners replacing a seasonal menu this month will book a twelve-image package at $160 when the formats, delivery date, and one revision round are explicit.
Test: make that genuine offer to five eligible owners you can appropriately approach. Track the exact offer and each response. Limit preparation to four hours; do not take money for work you cannot deliver.
Decision rule: one paid booking would justify testing delivery once. No booking would trigger review of the audience, offer clarity, and reasons volunteered. The rule does not claim a population conversion rate.
The number five and the one-booking rule are invented planning choices for this example. They are not recommended sample sizes, statistical confidence thresholds, or proof that a market exists.
Choose the risk that can stop the next step
Ask two questions: if the claim is wrong, does the plan materially change; and how much evidence do you already have? A new logo may be easy to test, but it is rarely the most consequential uncertainty when no one has accepted an offer.
A permission or feasibility issue can come first. If you cannot legally or reliably deliver the proposed service, testing a paid sale is premature. Resolve that constraint or choose a non-transactional discovery test with honest expectations.
Do not confuse “important” with “easy to measure.” Repeat demand may matter greatly but require time. Record it as unresolved rather than replacing it with a quick satisfaction question.
Record the result without rewriting the prediction
Keep the original statement and date alongside the actual observation. If an owner accepted only after you changed the package to $90, the result concerns the $90 offer, not the original $160 conditions. If only three of five eligible people received the proposal, record three offers, not five trials.
A result can be inconclusive. No response might reflect poor delivery of the message or bad timing rather than rejection of the price. A booking from a friend may involve relationship effects. Note these possibilities without inventing an explanation that rescues the idea.
Change one important element deliberately when possible. If audience, price, scope, and channel all change together, a better result will be harder to interpret.
Turn the result into a bounded next action
Continue when the evidence justifies the next limited commitment, change a claim when observations contradict it, pause when an essential constraint is unresolved, or stop when the experiment crosses its time or loss limit.
The validation canvas keeps hypothesis, test, evidence, and next decision in separate fields. Use the interview guide if you still need to understand the problem. Use the paid-pilot guide only when there is an offer you can honestly deliver.
A good hypothesis is useful even when it is contradicted. Its value is helping you avoid a larger unsupported commitment.
Your next useful step.
Open the Idea Validation Canvas, work through your own assumptions, and save a private copy as a JSON file.
Open the Idea Validation Canvas ↗Sources & scope
Consulted November 5, 2026. Sources support the definitions and methods identified in the text. Examples, exercises, and calculations are our own; they are not case studies or claims of customer outcomes.
General education, with US official resources where noted. This guide does not determine legal obligations, eligibility, or an individual financial decision. Read our editorial policy.